Michael Burry, the investor who famously predicted and profited from the 2008 US housing market crash, is now making a significant bet against the current artificial intelligence rally. Through his investment firm, Burry has taken short positions on two of the market's biggest AI champions: chipmaker Nvidia and data analytics firm Palantir.
The Core of Burry's Criticism: Accounting and Depreciation
In a detailed post on his Substack newsletter, Burry launched a fresh attack on Nvidia, dismissing the company's internal memo that attempted to rebut his initial concerns as "disingenuous on the face, and disappointing." He accused the world's most valuable company of deflecting his critique by creating "one straw man after another." Burry's primary concern is not with Nvidia's own financial practices but with the accounting choices of its largest customers.
Burry clarified that he never alleged Nvidia was manipulating its own depreciation schedules. "No one cares about Nvidia's own depreciation," he wrote, calling that a "straw man burnt." Instead, he zeroed in on the cloud giants—Microsoft, Meta, and Oracle—who are Nvidia's biggest buyers. Reports indicate these companies are extending the depreciation timelines for AI chips and servers from the standard 3 years to 5 or even 6 years.
This accounting move allows them to spread the massive cost of AI hardware over a longer period, which artificially inflates their current earnings by pushing future losses down the road. Burry sees this as a major red flag, suggesting that the reported profits in the AI sector are being propped up by assumptions that may not hold as technology rapidly evolves.
Echoes of the Dot-Com Bubble
Burry draws a direct parallel between the current AI frenzy and the dot-com bubble of 2000. He claims that today's surge is built on faster hardware, loftier assumptions, and bigger accounting risks. In his stark comparison, he identified Nvidia as the modern equivalent of Cisco, the networking giant that was at the heart of the tech euphoria before its dramatic crash.
"There is a Cisco at the center of it all," Burry wrote. "Its name is Nvidia." He backed this ominous warning with a specific forecast, predicting that tech companies could collectively understate their depreciation expenses by a staggering $176 billion between 2026 and 2028. He warned that companies like Meta and Oracle could be overstating their earnings by 20% to 27% if their depreciation timelines are not adjusted to reflect reality.
Nvidia's Response and Burry's Stance
Nvidia has publicly disputed Burry's analysis, particularly his figures related to the company's stock buyback program. The chipmaker clarified that it has repurchased $91 billion in shares since 2018, not the $112.5 billion Burry implied, stating he mistakenly included taxes related to Restricted Stock Units (RSUs).
The company also defended the longevity of its products, arguing that chips like the A100, released in 2020, are still in full production use, which they believe justifies longer depreciation periods for their customers.
Burry remains entirely unconvinced. "I stand by my analysis," he posted on X (formerly Twitter), adding that the full depth of his argument cannot be contained in a single social media post. He has disclosed that he continues to hold put options—a bet that the stock price will fall—on both Nvidia and Palantir. He clarified that these positions cost his fund about $10 million each, a figure far lower than the notional value might suggest.
He is now using his new paywalled newsletter, "Cassandra Unchained," to argue that the current AI investment cycle is as much about accounting decisions as it is about technological innovation. As investor debate intensifies over whether the AI boom has entered speculative territory, all eyes are on whether Michael Burry's bearish bet will once again prove prophetic.
